Segment size appears to influence

realtimeIngestionDelayMs

(metric that tracks consumer lag). That is when I decrease segment size, I've observed that the ingestion delay also decreases. I've also observed that ingestion delays only occur while the consuming segment is being sealed. Am I correct to assume this is occurring because ingestion is paused momentarily when the consuming segment is being sealed and there are less records to persist when the segments are smaller?

correct. the new segment won’t begin consumption until the previous segment is sealed